Research Agent MCP
by myeong-ga
use-gemini-chat.ts•9.15 kB
"use client"
import { useState, useEffect, useCallback, useRef } from "react"
import { useChat } from "@ai-sdk/react"
import type { CreateMessage, Message } from "@ai-sdk/react"
import type { Source, TokenUsage, ModelConfig } from "@/lib/types"
import { nanoid } from "@/lib/nanoid"
import type { SearchSuggestion } from "@/lib/types"
import { useLlmProvider } from "@/contexts/llm-provider-context"
import { getDefaultModelConfig } from "@/lib/models"
import { toast } from "sonner"
/*
aiMessages - ai sdk 에서 제공하는 메시지
optimisticMessages - 낙관적 업데이트를 위해 컴포넌트내에서 운용
messages - hook 에서 제공하며 ui component에서 사용
*/
export function useGeminiChat() {
const { getSelectedModel ,getReasoningType } = useLlmProvider()
const [sources, setSources] = useState<Source[]>([])
const [searchSuggestions, setSearchSuggestions] = useState<SearchSuggestion[]>([])
const [searchSuggestionsReasoning, setSearchSuggestionsReasoning] = useState<string>("")
const [searchSuggestionsConfidence, setSearchSuggestionsConfidence] = useState<number | null>(null)
const [optimisticMessages, setOptimisticMessages] = useState<Message[]>([])
const optimisticMessageIdRef = useRef<string | null>(null)
const [chatId, setChatId] = useState<string>(() => nanoid())
const [tokenUsage, setTokenUsage] = useState<TokenUsage | null>(null)
const [modelConfig, setModelConfig] = useState<ModelConfig>(() => {
const selectedModelId = getSelectedModel("gemini")
return getDefaultModelConfig(selectedModelId).config
})
const [responseModelConfig, setResponseModelConfig] = useState<ModelConfig | null>(null) // 마지막 응답에서 확인할 수 있는 model config , chatting 창 message display 와 함께 사용된다
const [responseSelectModel, setResponseSelectModel] = useState<string | null>(null) // 마지막 응답에서 확인할 수 있는 model config , chatting 창 message display 와 함께 사용된다
const [responseReasoningType, setResponseReasoningType] = useState<string | null>(null) // 마지막 응답에서 확인할 수 있는 model config , chatting 창 message display 와 함께 사용된다
const selectedModelId = getSelectedModel("gemini")
const reasoningType = getReasoningType("gemini", selectedModelId)
const {
messages: aiMessages,
status,
stop,
data,
setData,
append,
setMessages,
} = useChat({
api: "/api/chat/google",
id: chatId,
generateId: nanoid,
experimental_throttle: 50,
body: {
model: selectedModelId,
modelConfig,
reasoningType,
},
})
const messages = [
...optimisticMessages,
...aiMessages.filter((chatMsg) => !optimisticMessages.some((optMsg) => optMsg.id === chatMsg.id)),
]
useEffect(() => {
if (status === "streaming" && optimisticMessages.length > 0) {
setOptimisticMessages([])
optimisticMessageIdRef.current = null
}
}, [status, optimisticMessages])
useEffect(() => {
if (aiMessages.length > 0 && optimisticMessageIdRef.current) {
const hasOptimisticMessage = aiMessages.some((msg) => msg.id === optimisticMessageIdRef.current)
if (hasOptimisticMessage) {
setOptimisticMessages([])
optimisticMessageIdRef.current = null
}
}
}, [aiMessages])
useEffect(() => {
const selectedModelId = getSelectedModel("gemini")
const defaultConfig = getDefaultModelConfig(selectedModelId).config
setModelConfig(defaultConfig)
}, [getSelectedModel])
useEffect(() => {
if (data && Array.isArray(data)) {
for (const item of data) {
if (item && typeof item === "object" && "type" in item) {
if (item.type === "sources" && "sources" in item && Array.isArray(item.sources)) {
setSources(item.sources as Source[])
} else if (item.type === "searchSuggestions") {
if ("searchSuggestions" in item && Array.isArray(item.searchSuggestions)) {
const suggestions = item.searchSuggestions.map((term) => ({ term }))
setSearchSuggestions(suggestions as SearchSuggestion[])
if ("confidence" in item && typeof item.confidence === "number") {
setSearchSuggestionsConfidence(item.confidence)
}
if ("reasoning" in item && typeof item.reasoning === "string") {
setSearchSuggestionsReasoning(item.reasoning)
}
}
} else if (item.type === "usage" && typeof item.usage === "object") {
const usage = item.usage as any
if (
typeof usage.promptTokens === "number" &&
typeof usage.completionTokens === "number" &&
typeof usage.totalTokens === "number"
) {
setTokenUsage({
promptTokens: usage.promptTokens,
completionTokens: usage.completionTokens,
totalTokens: usage.totalTokens,
finishReason: usage.finishReason || "unknown",
})
}
} else if (item.type === "model-config" && typeof item.config === "object") {
// Handle model config response from server
const config = item.config as unknown as ModelConfig
if (
typeof config.temperature === "number" &&
typeof config.topP === "number" &&
typeof config.maxTokens === "number"
) {
setResponseModelConfig(config)
}
} else if (item.type === "selected-model" && typeof item.model === "string") {
const responseModel = item.model as string
setResponseSelectModel(responseModel)
} else if (item.type === "reasoning-type" && typeof item.reasoning === "string") {
const responseReasoning = item.reasoning as string
setResponseReasoningType(responseReasoning)
} else if (item.type === "cleaned-text" && "text" in item && typeof item.text === "string") {
setMessages((prevMessages) => {
const newMessages = [...prevMessages]
for (let i = newMessages.length - 1; i >= 0; i--) {
if (newMessages[i].role === "assistant") {
newMessages[i] = {
...newMessages[i],
content: item.text as string,
}
break
}
}
return newMessages
})
}
}
}
}
}, [data, setMessages])
const updateModelConfig = useCallback((config: Partial<ModelConfig>, showToast = false) => {
setModelConfig((prevConfig) => {
const newConfig = {
...prevConfig,
...config,
}
if (showToast) {
toast.success("Model configuration updated", {
description: `Temperature: ${newConfig.temperature.toFixed(2)}, Max Tokens: ${newConfig.maxTokens}`,
duration: 3000,
})
}
return newConfig
})
}, [])
const sendMessage = async (message: string ) => {
setData(undefined);
setSources([])
setSearchSuggestions([])
setSearchSuggestionsReasoning("")
setSearchSuggestionsConfidence(null)
setTokenUsage(null)
setResponseModelConfig(null)
const userMessage: Message = { id: nanoid(), role: "user", content: message }
// typeof message === "string" ? { id: nanoid(), role: "user", content: message } : { id: nanoid(), ...message }
// 낙관적 업데이트: 사용자 메시지를 낙관적 메시지 배열에 추가
setOptimisticMessages([userMessage])
optimisticMessageIdRef.current = userMessage.id
try {
await append(
{
role: "user",
content: userMessage.content,
id: userMessage.id,
},
{
body: {
model: selectedModelId,
modelConfig,
reasoningType,
},
},
)
} catch (error) {
console.error("Failed to send message:", error)
setOptimisticMessages([])
optimisticMessageIdRef.current = null
}
}
const resetChat = useCallback(() => {
setOptimisticMessages([])
setSources([])
setSearchSuggestions([])
setSearchSuggestionsReasoning("")
setSearchSuggestionsConfidence(null)
setMessages([])
setData(undefined);
setChatId(nanoid())
setTokenUsage(null)
setResponseModelConfig(null)
// Reset model config to default
const selectedModelId = getSelectedModel("gemini")
setModelConfig(getDefaultModelConfig(selectedModelId).config)
}, [setMessages, getSelectedModel])
return {
messages,
status,
stop,
tokenUsage,
sources,
searchSuggestions,
searchSuggestionsReasoning,
searchSuggestionsConfidence,
sendMessage,
chatId,
resetChat,
modelConfig, // LLM 모델 파라미터
responseModelConfig,
responseSelectModel,
responseReasoningType,
updateModelConfig,
}
}
Latest Blog Posts
- Federated Learning with MCP: Building Privacy-Preserving Agents Across Distributed EdgesBy Om-Shree-0709 on .SecuremcpLearning
- MCP Moves to the Linux Foundation: Neutral Stewardship for Agentic InfrastructureBy Om-Shree-0709 on .mcpanthropicLinux Foundation
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/myeong-ga/research-agent-mcp-0.36-pro-preview-06-01'
If you have feedback or need assistance with the MCP directory API, please join our Discord server